| 摘要: | 隨著工業4.0及人機協作技術的快速發展，對微小精密元件的高精度三維檢測需求日益迫切。傳統二維機器視覺技術僅能針對特定平面進行檢測，難以滿足複雜幾何結構的全方位量測需求。本研究開發一套基於空間光調制器(Spatial Light Modulator, SLM)之結構光三維形貌重建系統，並提出互補影像對比中值法(CIMT)作為格雷碼二值化解決方案。
本系統採用正投影斜收光配置，利用SLM取代傳統數位微鏡元件(Digital Micromirror Device, DMD)進行結構光投影。相較於DMD採用脈衝寬度調制(Pulse-width modulation , PWM)技術，SLM運用偏振調制原理具備更高的灰階準確性與穩定性。系統結合六位元拓展格雷碼與四步移相技術，透過伽瑪校正(Gamma Correction)確保投影條紋具有理想弦波波形。收光端引入沙姆攝影原理(Scheimpflug Principle)，解決斜向拍攝之景深限制問題，實現全景深清晰成像。
針對格雷碼在複雜表面條件下的二值化難題，本研究創新提出(Complementary Image Median Thresholding , CIMT)方法，利用格雷碼方波條紋的特性，投影一正一反互補影像，為每個像素計算專屬自適應閾值，並增加陰影區域判別機制，當正反影像灰階值差異小於設定閾值時，則判定為陰影區域並標記為未定義，避免生成錯誤相位及點雲。有效改善傳統固定閾值法、Otsu二值化及自適應閾值法在非均勻反射率、曲率變化表面或光照不均勻條件下的局限性。實驗結果顯示，CIMT方法在處理電子連接器等多材質複合組裝之物體時表現出優異的穩定性與準確性。
系統性能分析顯示，在11mm×9mm視野範圍內可達到18μm的亞像素測量精度，成功實現對鐵質、塑膠、黃銅等不同反射特性材質的精確的三維形貌重建。本研究為非接觸式精密檢測技術提供新的解決方案，具有重要的學術價值與工業應用潛力。
;With the rapid development of Industry 4.0 and human-machine collaboration technologies, there is an increasingly urgent demand for high-precision three-dimensional inspection of miniature precision components. Traditional two-dimensional machine vision technology can only perform inspection on specific planes, making it difficult to meet the comprehensive measurement requirements of complex geometric structures. This study develops a structured light three-dimensional topography reconstruction system based on Spatial Light Modulator (SLM) and proposes the Complementary Image Median Thresholding (CIMT) method as a Gray code binarization solution.
The system adopts an orthogonal projection oblique collection configuration, utilizing SLM to replace traditional Digital Micromirror Device (DMD) for structured light projection. Compared to DMD′s Pulse-width Modulation (PWM) technology, SLM employs polarization modulation principles with higher grayscale accuracy and stability. The system combines six-bit extended Gray code with four-step phase shifting technology, ensuring ideal sinusoidal waveform through Gamma Correction. The collection end incorporates the Scheimpflug Principle to solve depth of field limitations in oblique imaging.
Addressing Gray code binarization challenges under complex surface conditions, this study innovatively proposes the CIMT method. By projecting complementary positive and negative images, it calculates adaptive thresholds for each pixel and adds shadow region discrimination. When grayscale differences between positive and negative images fall below the threshold, regions are marked as undefined, avoiding erroneous phases and point clouds. This effectively improves limitations of traditional fixed threshold, Otsu binarization, and adaptive threshold methods under non-uniform reflectivity or uneven illumination conditions.
System performance analysis shows sub-pixel measurement accuracy of 18μm within an 11mm×9mm field of view, successfully achieving precise three-dimensional topography reconstruction of materials with different reflective characteristics such as iron, plastic, and brass. This study provides a new solution for non-contact precision inspection technology with significant academic value and industrial application potential. |
